I'm converting a Shapely MultiPolygon to a PatchCollection, and first colouring each Polygon like so:

# ldn_mp is a MultiPolygon
cm = plt.get_cmap('RdBu')
num_colours = len(ldn_mp)fig = plt.figure()
ax = fig.add_subplot(111)
minx, miny, maxx, maxy = ldn_mp.bounds
w, h = maxx - minx, maxy - miny
ax.set_xlim(minx - 0.2 * w, maxx + 0.2 * w)
ax.set_ylim(miny - 0.2 * h, maxy + 0.2 * h)
ax.set_aspect(1)patches = []
for poly in ldn_mp:colour = cm(1. * len(filter(poly.contains, points)) / num_colours)patches.append(PolygonPatch(poly, fc=colour, ec='#555555', lw=0.2, alpha=1., zorder=1))
pc = PatchCollection(patches, match_original=True)
ax.add_collection(pc)
ax.set_xticks([])
ax.set_yticks([])
plt.title("Density of NO$^2$ Sensors by Borough")
plt.tight_layout()
plt.show()

But I'd like to add a colorbar to my plot, based upon the PatchCollection colors. I'm not sure how to go about that; do I pass the cmap keyword when creating pc? How do I then call set_array() with the colours I've used?

I had the same problem a little while ago. For each polygon I saved the corresponding color to a list named mycolors:

mycolors=[]
...
mycolors.append(SSTvalue)
path_patch = patches.PathPatch(mypath, lw=1)
mypatches.append(path_patch)

I looped over a series of multipolygons stored in a Shapefile and stored each patch in a collection. After that I plotted the polygons using the color information I had stored in the list, which was converted to an array eventually, and added a colorbar:

p = PatchCollection(mypatches, cmap=plt.get_cmap('RdYlBu_r'), alpha=1.0)
p.set_array(array(mycolors))
p.set_clim([np.ma.min(mycolors),np.ma.max(mycolors)])
plt.colorbar(p,shrink=0.5)

The full script I used to plot temperature values with colors and a colorbar for large marine ecosystems of the world represented by polygons can be found here. Hope this helps. Cheers, Trond